platform/kernel/linux-rpi.git
4 years agoefi/efivars: Add missing kobject_put() in sysfs entry creation error path
Ard Biesheuvel [Fri, 22 May 2020 16:15:49 +0000 (18:15 +0200)]
efi/efivars: Add missing kobject_put() in sysfs entry creation error path

The documentation provided by kobject_init_and_add() clearly spells out
the need to call kobject_put() on the kobject if an error is returned.
Add this missing call to the error path.

Cc: <stable@vger.kernel.org>
Reported-by: 亿一 <teroincn@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Use pool allocation for the command line
Ard Biesheuvel [Tue, 19 May 2020 08:43:01 +0000 (10:43 +0200)]
efi/libstub: Use pool allocation for the command line

Now that we removed the memory limit for the allocation of the
command line, there is no longer a need to use the page based
allocator so switch to a pool allocation instead.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Don't parse overlong command lines
Arvind Sankar [Thu, 21 May 2020 00:29:21 +0000 (20:29 -0400)]
efi/libstub: Don't parse overlong command lines

Check if the command line passed in is larger than COMMAND_LINE_SIZE,
and truncate it to the last full argument if so.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200521002921.69650-1-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Use snprintf with %ls to convert the command line
Arvind Sankar [Mon, 18 May 2020 19:07:16 +0000 (15:07 -0400)]
efi/libstub: Use snprintf with %ls to convert the command line

Now we can use snprintf to do the UTF-16 to UTF-8 translation for the
command line.

Drop the special "zero" trick to handle an empty command line. This was
unnecessary even before this since with options_chars == 0,
efi_utf16_to_utf8 would not have accessed options at all. snprintf won't
access it either with a precision of 0.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-25-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Get the exact UTF-8 length
Arvind Sankar [Mon, 18 May 2020 19:07:15 +0000 (15:07 -0400)]
efi/libstub: Get the exact UTF-8 length

efi_convert_cmdline currently overestimates the length of the equivalent
UTF-8 encoding.

snprintf can now be used to do the conversion to UTF-8, however, it does
not have a way to specify the size of the UTF-16 string, only the size
of the resulting UTF-8 string. So in order to use it, we need to
precalculate the exact UTF-8 size.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-24-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Use %ls for filename
Arvind Sankar [Mon, 18 May 2020 19:07:14 +0000 (15:07 -0400)]
efi/libstub: Use %ls for filename

efi_printk can now handle the UTF-16 filename, so print it using efi_err
instead of a separate efi_char16_puts call.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-23-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Add UTF-8 decoding to efi_puts
Arvind Sankar [Mon, 18 May 2020 19:07:13 +0000 (15:07 -0400)]
efi/libstub: Add UTF-8 decoding to efi_puts

In order to be able to use the UTF-16 support added to vsprintf in the
previous commit, enhance efi_puts to decode UTF-8 into UTF-16. Invalid
UTF-8 encodings are passed through unchanged.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-22-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Add support for wchar_t (UTF-16)
Arvind Sankar [Mon, 18 May 2020 19:07:12 +0000 (15:07 -0400)]
efi/printf: Add support for wchar_t (UTF-16)

Support %lc and %ls to output UTF-16 strings (converted to UTF-8).

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-21-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Add an option to list out the available GOP modes
Arvind Sankar [Mon, 18 May 2020 19:07:11 +0000 (15:07 -0400)]
efi/gop: Add an option to list out the available GOP modes

Add video=efifb:list option to list the modes that are available.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-20-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Add definitions for console input and events
Arvind Sankar [Mon, 18 May 2020 19:07:10 +0000 (15:07 -0400)]
efi/libstub: Add definitions for console input and events

Add the required typedefs etc for using con_in's simple text input
protocol, and for using the boottime event services.

Also add the prototype for the "stall" boot service.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-19-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Implement printk-style logging
Arvind Sankar [Wed, 20 May 2020 17:07:54 +0000 (19:07 +0200)]
efi/libstub: Implement printk-style logging

Use the efi_printk function in efi_info/efi_err, and add efi_debug. This
allows formatted output at different log levels.

Add the notion of a loglevel instead of just quiet/not-quiet, and
parse the efi=debug kernel parameter in addition to quiet.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200520170223.GA3333632@rani.riverdale.lan/
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Turn vsprintf into vsnprintf
Arvind Sankar [Mon, 18 May 2020 19:07:08 +0000 (15:07 -0400)]
efi/printf: Turn vsprintf into vsnprintf

Implement vsnprintf instead of vsprintf to avoid the possibility of a
buffer overflow.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-17-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Abort on invalid format
Arvind Sankar [Mon, 18 May 2020 19:07:07 +0000 (15:07 -0400)]
efi/printf: Abort on invalid format

If we get an invalid conversion specifier, bail out instead of trying to
fix it up. The format string likely has a typo or assumed we support
something that we don't, in either case the remaining arguments won't
match up with the remaining format string.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-16-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Refactor code to consolidate padding and output
Arvind Sankar [Mon, 18 May 2020 19:07:06 +0000 (15:07 -0400)]
efi/printf: Refactor code to consolidate padding and output

Consolidate the actual output of the formatted text into one place.

Fix a couple of edge cases:
1. If 0 is printed with a precision of 0, the printf specification says
   that nothing should be output, with one exception (2b).
2. The specification for octal alternate format (%#o) adds the leading
   zero not as a prefix as the 0x for hexadecimal is, but by increasing
   the precision if necessary to add the zero. This means that
   a. %#.2o turns 8 into "010", but 1 into "01" rather than "001".
   b. %#.o prints 0 as "0" rather than "", unlike the situation for
      decimal, hexadecimal and regular octal format, which all output an
      empty string.

Reduce the space allocated for printing a number to the maximum actually
required (22 bytes for a 64-bit number in octal), instead of the 66
bytes previously allocated.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-15-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Handle null string input
Arvind Sankar [Mon, 18 May 2020 19:07:05 +0000 (15:07 -0400)]
efi/printf: Handle null string input

Print "(null)" for 's' if the input is a NULL pointer.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-14-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Factor out integer argument retrieval
Arvind Sankar [Mon, 18 May 2020 19:07:04 +0000 (15:07 -0400)]
efi/printf: Factor out integer argument retrieval

Factor out the code to get the correct type of numeric argument into a
helper function.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-13-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Factor out width/precision parsing
Arvind Sankar [Mon, 18 May 2020 19:07:03 +0000 (15:07 -0400)]
efi/printf: Factor out width/precision parsing

Factor out the width/precision parsing into a helper function.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-12-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Merge 'p' with the integer formats
Arvind Sankar [Mon, 18 May 2020 19:07:02 +0000 (15:07 -0400)]
efi/printf: Merge 'p' with the integer formats

Treat 'p' as a hexadecimal integer with precision equal to the number of
digits in void *.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-11-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Fix minor bug in precision handling
Arvind Sankar [Mon, 18 May 2020 19:07:01 +0000 (15:07 -0400)]
efi/printf: Fix minor bug in precision handling

A negative precision should be ignored completely, and the presence of a
valid precision should turn off the 0 flag.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-10-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Factor out flags parsing and handle '%' earlier
Arvind Sankar [Mon, 18 May 2020 19:07:00 +0000 (15:07 -0400)]
efi/printf: Factor out flags parsing and handle '%' earlier

Move flags parsing code out into a helper function.

The '%%' case can be handled up front: it is not allowed to have flags,
width etc.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-9-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Add 64-bit and 8-bit integer support
Arvind Sankar [Mon, 18 May 2020 19:06:59 +0000 (15:06 -0400)]
efi/printf: Add 64-bit and 8-bit integer support

Support 'll' qualifier for long long by copying the decimal printing
code from lib/vsprintf.c. For simplicity, the 32-bit code is used on
64-bit architectures as well.

Support 'hh' qualifier for signed/unsigned char type integers.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-8-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/printf: Drop %n format and L qualifier
Arvind Sankar [Mon, 18 May 2020 19:06:58 +0000 (15:06 -0400)]
efi/printf: Drop %n format and L qualifier

%n is unused and deprecated.

The L qualifer is parsed but not actually implemented.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-7-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Optimize for size instead of speed
Arvind Sankar [Mon, 18 May 2020 19:06:57 +0000 (15:06 -0400)]
efi/libstub: Optimize for size instead of speed

Reclaim the bloat from the addition of printf by optimizing the stub for
size. With gcc 9, the text size of the stub is:

ARCH    before  +printf    -Os
arm      35197    37889  34638
arm64    34883    38159  34479
i386     18571    21657  17025
x86_64   25677    29328  22144

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-6-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Add a basic printf implementation
Arvind Sankar [Mon, 18 May 2020 19:06:56 +0000 (15:06 -0400)]
efi/libstub: Add a basic printf implementation

Copy vsprintf from arch/x86/boot/printf.c to get a simple printf
implementation.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-5-nivedita@alum.mit.edu
[ardb: add some missing braces in if...else clauses]
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Buffer output of efi_puts
Arvind Sankar [Mon, 18 May 2020 19:06:55 +0000 (15:06 -0400)]
efi/libstub: Buffer output of efi_puts

Use a buffer to convert the string to UTF-16. This will reduce the
number of firmware calls required to print the string from one per
character to one per string in most cases.

Cast the input char to unsigned char before converting to efi_char16_t
to avoid sign-extension in case there are any non-ASCII characters in
the input.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-4-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Rename efi_[char16_]printk to efi_[char16_]puts
Arvind Sankar [Mon, 18 May 2020 19:06:54 +0000 (15:06 -0400)]
efi/libstub: Rename efi_[char16_]printk to efi_[char16_]puts

These functions do not support formatting, unlike printk. Rename them to
puts to make that clear.

Move the implementations of these two functions next to each other.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-3-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Include dependencies of efistub.h
Arvind Sankar [Mon, 18 May 2020 19:06:53 +0000 (15:06 -0400)]
efi/libstub: Include dependencies of efistub.h

Add #include directives for include files that efistub.h depends on,
instead of relying on them having been included by the C source files
prior to efistub.h.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200518190716.751506-2-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Correct comment typos
Joe Perches [Thu, 30 Apr 2020 19:12:42 +0000 (12:12 -0700)]
efi/libstub: Correct comment typos

Fix a couple typos in comments.

Signed-off-by: Joe Perches <joe@perches.com>
Link: https://lore.kernel.org/r/ec53e67b3ac928922807db3cb1585e911971dadc.1588273612.git.joe@perches.com
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Make efi_printk() input argument const char*
Ard Biesheuvel [Tue, 5 May 2020 08:27:18 +0000 (10:27 +0200)]
efi/libstub: Make efi_printk() input argument const char*

To help the compiler figure out that efi_printk() will not modify
the string it is given, make the input argument type const char*.

While at it, simplify the implementation as well.

Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/x86: Work around LLVM ELF quirk build regression
Ard Biesheuvel [Mon, 4 May 2020 08:06:29 +0000 (10:06 +0200)]
efi/libstub/x86: Work around LLVM ELF quirk build regression

When building the x86 EFI stub with Clang, the libstub Makefile rules
that manipulate the ELF object files may throw an error like:

    STUBCPY drivers/firmware/efi/libstub/efi-stub-helper.stub.o
  strip: drivers/firmware/efi/libstub/efi-stub-helper.stub.o: Failed to find link section for section 10
  objcopy: drivers/firmware/efi/libstub/efi-stub-helper.stub.o: Failed to find link section for section 10

This is the result of a LLVM feature [0] where symbol references are
stored in a LLVM specific .llvm_addrsig section in a non-transparent way,
causing generic ELF tools such as strip or objcopy to choke on them.

So force the compiler not to emit these sections, by passing the
appropriate command line option.

[0] https://sourceware.org/bugzilla/show_bug.cgi?id=23817

Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Peter Collingbourne <pcc@google.com>
Cc: Sami Tolvanen <samitolvanen@google.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Suggested-by: Fangrui Song <maskray@google.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Fix mixed mode boot issue after macro refactor
Arvind Sankar [Mon, 4 May 2020 15:02:48 +0000 (11:02 -0400)]
efi/libstub: Fix mixed mode boot issue after macro refactor

Commit

  22090f84bc3f ("efi/libstub: unify EFI call wrappers for non-x86")

refactored the macros that are used to provide wrappers for mixed-mode
calls on x86, allowing us to boot a 64-bit kernel on 32-bit firmware.

Unfortunately, this broke mixed mode boot due to the fact that
efi_is_native() is not a macro on x86.

All of these macros should go together, so rather than testing each one
to see if it is defined, condition the generic macro definitions on a
new ARCH_HAS_EFISTUB_WRAPPERS, and remove the wrapper definitions on x86
as well if CONFIG_EFI_MIXED is not enabled.

Fixes: 22090f84bc3f ("efi/libstub: unify EFI call wrappers for non-x86")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200504150248.62482-1-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Check return value of efi_parse_options
Arvind Sankar [Thu, 30 Apr 2020 18:28:43 +0000 (14:28 -0400)]
efi/libstub: Check return value of efi_parse_options

efi_parse_options can fail if it is unable to allocate space for a copy
of the command line. Check the return value to make sure it succeeded.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-12-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/x86: Support builtin command line
Arvind Sankar [Thu, 30 Apr 2020 18:28:42 +0000 (14:28 -0400)]
efi/x86: Support builtin command line

Add support for the x86 CMDLINE_BOOL and CMDLINE_OVERRIDE configuration
options.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-11-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Unify initrd loading across architectures
Arvind Sankar [Thu, 30 Apr 2020 18:28:41 +0000 (14:28 -0400)]
efi/libstub: Unify initrd loading across architectures

Factor out the initrd loading into a common function that can be called
both from the generic efi-stub.c and the x86-specific x86-stub.c.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-10-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/x86: Move command-line initrd loading to efi_main
Arvind Sankar [Thu, 30 Apr 2020 18:28:40 +0000 (14:28 -0400)]
efi/x86: Move command-line initrd loading to efi_main

Consolidate the initrd loading in efi_main.

The command line options now need to be parsed only once.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-9-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Upgrade ignored dtb= argument message to error
Arvind Sankar [Thu, 30 Apr 2020 18:28:39 +0000 (14:28 -0400)]
efi/libstub: Upgrade ignored dtb= argument message to error

Use efi_err if we ignore a command-line dtb= argument, so that it shows
up even on a quiet boot.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-8-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/tpm: Use efi_err for error messages
Arvind Sankar [Thu, 30 Apr 2020 18:28:38 +0000 (14:28 -0400)]
efi/tpm: Use efi_err for error messages

Use efi_err instead of bare efi_printk for error messages.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-7-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Use efi_err for error messages
Arvind Sankar [Thu, 30 Apr 2020 18:28:37 +0000 (14:28 -0400)]
efi/gop: Use efi_err for error messages

Use efi_err instead of bare efi_printk for error messages.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-6-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/x86: Use efi_err for error messages
Arvind Sankar [Thu, 30 Apr 2020 18:28:36 +0000 (14:28 -0400)]
efi/x86: Use efi_err for error messages

Use efi_err instead of bare efi_printk for error messages.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-5-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Move pr_efi/pr_efi_err into efi namespace
Arvind Sankar [Thu, 30 Apr 2020 18:28:35 +0000 (14:28 -0400)]
efi/libstub: Move pr_efi/pr_efi_err into efi namespace

Rename pr_efi to efi_info and pr_efi_err to efi_err to make it more
obvious that they are part of the EFI stub and not generic printk infra.

Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-4-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Add a helper function to split 64-bit values
Arvind Sankar [Thu, 30 Apr 2020 18:28:34 +0000 (14:28 -0400)]
efi/libstub: Add a helper function to split 64-bit values

In several places 64-bit values need to be split up into two 32-bit
fields, in order to be backward-compatible with the old 32-bit ABIs.

Instead of open-coding this, add a helper function to set a 64-bit value
as two 32-bit fields.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-3-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/x86: Use correct size for boot_params
Arvind Sankar [Thu, 30 Apr 2020 18:28:33 +0000 (14:28 -0400)]
efi/x86: Use correct size for boot_params

struct boot_params is only 4096 bytes, not 16384. Fix this by using
sizeof(struct boot_params) instead of hardcoding the incorrect value.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200430182843.2510180-2-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Re-enable command line initrd loading for x86
Ard Biesheuvel [Sat, 25 Apr 2020 10:22:04 +0000 (12:22 +0200)]
efi/libstub: Re-enable command line initrd loading for x86

Commit:

  cf6b83664895a5 ("efi/libstub: Make initrd file loader configurable")

inadvertently disabled support on x86 for loading an initrd passed via
the initrd= option on the kernel command line.

Add X86 to the newly introduced Kconfig option's title and depends
declarations, so it gets enabled by default, as before.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: linux-efi@vger.kernel.org
4 years agoMerge tag 'efi-next' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi into...
Ingo Molnar [Sat, 25 Apr 2020 08:25:02 +0000 (10:25 +0200)]
Merge tag 'efi-next' of git://git./linux/kernel/git/efi/efi into efi/core

Pull EFI changes for v5.8 from Ard Biesheuvel:

"- preliminary changes for RISC-V
 - add support for setting the resolution on the EFI framebuffer
 - simplify kernel image loading for arm64
 - Move .bss into .data via the linker script instead of relying on symbol
   annotations.
 - Get rid of __pure getters to access global variables
 - Clean up the config table matching arrays"

Signed-off-by: Ingo Molnar <mingo@kernel.org>
4 years agoefi: Move arch_tables check to caller
Ard Biesheuvel [Thu, 26 Mar 2020 08:34:35 +0000 (09:34 +0100)]
efi: Move arch_tables check to caller

Instead of making match_config_table() test its table_types pointer for
NULL-ness, omit the call entirely if no arch_tables pointer was provided
to efi_config_parse_tables().

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi: Clean up config table description arrays
Ard Biesheuvel [Thu, 26 Mar 2020 08:24:14 +0000 (09:24 +0100)]
efi: Clean up config table description arrays

Increase legibility by adding whitespace to the efi_config_table_type_t
arrays that describe which EFI config tables we look for when going over
the firmware provided list. While at it, replace the 'name' char pointer
with a char array, which is more space efficient on relocatable 64-bit
kernels, as it avoids a 8 byte pointer and the associated relocation
data (24 bytes when using RELA format)

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/x86: Avoid getter function for efi_is64
Ard Biesheuvel [Thu, 16 Apr 2020 19:15:23 +0000 (21:15 +0200)]
efi/libstub/x86: Avoid getter function for efi_is64

We no longer need to take special care when using global variables
in the EFI stub, so switch to a simple symbol reference for efi_is64.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Drop __pure getters for EFI stub options
Ard Biesheuvel [Thu, 16 Apr 2020 16:45:24 +0000 (18:45 +0200)]
efi/libstub: Drop __pure getters for EFI stub options

The practice of using __pure getter functions to access global
variables in the EFI stub dates back to the time when we had to
carefully prevent GOT entries from being emitted, because we
could not rely on the toolchain to do this for us.

Today, we use the hidden visibility pragma for all EFI stub source
files, which now all live in the same subdirectory, and we apply a
sanity check on the objects, so we can get rid of these getter
functions and simply refer to global data objects directly.

So switch over the remaining boolean variables carrying options set
on the kernel command line.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Drop __pure getter for efi_system_table
Ard Biesheuvel [Thu, 16 Apr 2020 16:38:06 +0000 (18:38 +0200)]
efi/libstub: Drop __pure getter for efi_system_table

The practice of using __pure getter functions to access global
variables in the EFI stub dates back to the time when we had to
carefully prevent GOT entries from being emitted, because we
could not rely on the toolchain to do this for us.

Today, we use the hidden visibility pragma for all EFI stub source
files, which now all live in the same subdirectory, and we apply a
sanity check on the objects, so we can get rid of these getter
functions and simply refer to global data objects directly.

Start with efi_system_table(), and convert it into a global variable.
While at it, make it a pointer-to-const, because we can.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi: Kill __efistub_global
Arvind Sankar [Thu, 16 Apr 2020 15:12:27 +0000 (11:12 -0400)]
efi: Kill __efistub_global

Now that both arm and x86 are using the linker script to place the EFI
stub's global variables in the correct section, remove __efistub_global.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20200416151227.3360778-4-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/x86: Remove __efistub_global and add relocation check
Arvind Sankar [Thu, 16 Apr 2020 15:12:26 +0000 (11:12 -0400)]
efi/x86: Remove __efistub_global and add relocation check

Instead of using __efistub_global to force variables into the .data
section, leave them in the .bss but pull the EFI stub's .bss section
into .data in the linker script for the compressed kernel.

Add relocation checking for x86 as well to catch non-PC-relative
relocations that require runtime processing, since the EFI stub does not
do any runtime relocation processing.

This will catch, for example, data relocations created by static
initializers of pointers.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200416151227.3360778-3-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/arm: Remove __efistub_global annotation
Arvind Sankar [Thu, 16 Apr 2020 15:12:25 +0000 (11:12 -0400)]
efi/arm: Remove __efistub_global annotation

Instead of using __efistub_global to force variables into the .data
section, leave them in the .bss but pull the EFI stub's .bss section
into .data in the linker script for the compressed kernel.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20200416151227.3360778-2-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Move efi_relocate_kernel() into separate source file
Ard Biesheuvel [Mon, 13 Apr 2020 12:23:57 +0000 (14:23 +0200)]
efi/libstub: Move efi_relocate_kernel() into separate source file

Move efi_relocate_kernel() into a separate source file, so that it
only gets pulled into builds for architectures that use it. Since
efi_relocate_kernel() is the only user of efi_low_alloc(), let's
move that over as well.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/arm64: Switch to ordinary page allocator for kernel image
Ard Biesheuvel [Tue, 31 Mar 2020 08:59:39 +0000 (10:59 +0200)]
efi/libstub/arm64: Switch to ordinary page allocator for kernel image

It is no longer necessary to locate the kernel as low as possible in
physical memory, and so we can switch from efi_low_alloc() [which is
a rather nasty concoction on top of GetMemoryMap()] to a new helper
called efi_allocate_pages_aligned(), which simply rounds up the size
to account for the alignment, and frees the misaligned pages again.

So considering that the kernel can live anywhere in the physical
address space, as long as its alignment requirements are met, let's
switch to efi_allocate_pages_aligned() to allocate the pages.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Add API function to allocate aligned memory
Ard Biesheuvel [Fri, 27 Mar 2020 15:09:40 +0000 (16:09 +0100)]
efi/libstub: Add API function to allocate aligned memory

Break out the code to create an aligned page allocation from mem.c
and move it into a function efi_allocate_pages_aligned() in alignedmem.c.
Update efi_allocate_pages() to invoke it unless the minimum alignment
equals the EFI page size (4 KB), in which case the ordinary page
allocator is sufficient. This way, efi_allocate_pages_aligned() will
only be pulled into the build if it is actually being used (which will
be on arm64 only in the immediate future)

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/arm64: Simplify randomized loading of kernel image
Ard Biesheuvel [Mon, 13 Apr 2020 13:36:37 +0000 (15:36 +0200)]
efi/libstub/arm64: Simplify randomized loading of kernel image

The KASLR code path in the arm64 version of the EFI stub incorporates
some overly complicated logic to randomly allocate a region of the right
alignment: there is no need to randomize the placement of the kernel
modulo 2 MiB separately from the placement of the 2 MiB aligned allocation
itself - we can simply follow the same logic used by the non-randomized
placement, which is to allocate at the correct alignment, and only take
TEXT_OFFSET into account if it is not a round multiple of the alignment.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/arm64: Replace 'preferred' offset with alignment check
Ard Biesheuvel [Fri, 27 Mar 2020 16:23:52 +0000 (17:23 +0100)]
efi/libstub/arm64: Replace 'preferred' offset with alignment check

The notion of a 'preferred' load offset for the kernel dates back to the
times when the kernel's primary mapping overlapped with the linear region,
and memory below it could not be used at all.

Today, the arm64 kernel does not really care where it is loaded in physical
memory, as long as the alignment requirements are met, and so there is no
point in unconditionally moving the kernel to a new location in memory at
boot. Instead, we can
- check for a KASLR seed, and randomly reallocate the kernel if one is
  provided
- otherwise, check whether the alignment requirements are met for the
  current placement of the kernel, and just run it in place if they are
- finally, do an ordinary page allocation and reallocate the kernel to a
  suitably aligned buffer anywhere in memory.

By the same reasoning, there is no need to take TEXT_OFFSET into account
if it is a round multiple of the minimum alignment, which is the usual
case for relocatable kernels with TEXT_OFFSET randomization disabled.
Otherwise, it suffices to use the relative misaligment of TEXT_OFFSET
when reallocating the kernel.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/random: Increase random alloc granularity
Ard Biesheuvel [Mon, 13 Apr 2020 14:06:56 +0000 (16:06 +0200)]
efi/libstub/random: Increase random alloc granularity

The implementation of efi_random_alloc() arbitrarily truncates the
provided random seed to 16 bits, which limits the granularity of the
randomly chosen allocation offset in memory. This is currently only
an issue if the size of physical memory exceeds 128 GB, but going
forward, we will reduce the allocation alignment to 64 KB, and this
means we need to increase the granularity to ensure that the random
memory allocations are distributed evenly.

We will need to switch to 64-bit arithmetic for the multiplication,
but this does not result in 64-bit integer intrinsic calls on ARM or
on i386.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/random: Align allocate size to EFI_ALLOC_ALIGN
Ard Biesheuvel [Mon, 13 Apr 2020 14:01:05 +0000 (16:01 +0200)]
efi/libstub/random: Align allocate size to EFI_ALLOC_ALIGN

The EFI stub uses a per-architecture #define for the minimum base
and size alignment of page allocations, which is set to 4 KB for
all architecures except arm64, which uses 64 KB, to ensure that
allocations can always be (un)mapped efficiently, regardless of
the page size used by the kernel proper, which could be a kexec'ee

The API wrappers around page based allocations assume that this
alignment is always taken into account, and so efi_free() will
also round up its size argument to EFI_ALLOC_ALIGN.

Currently, efi_random_alloc() does not honour this alignment for
the allocated size, and so freeing such an allocation may result
in unrelated memory to be freed, potentially leading to issues
after boot. So let's round up size in efi_random_alloc() as well.

Fixes: 2ddbfc81eac84a29 ("efi: stub: add implementation of efi_random_alloc()")
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Allow automatically choosing the best mode
Arvind Sankar [Sat, 28 Mar 2020 16:06:01 +0000 (12:06 -0400)]
efi/gop: Allow automatically choosing the best mode

Add the ability to automatically pick the highest resolution video mode
(defined as the product of vertical and horizontal resolution) by using
a command-line argument of the form
video=efifb:auto

If there are multiple modes with the highest resolution, pick one with
the highest color depth.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200328160601.378299-2-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Allow specifying depth as well as resolution
Arvind Sankar [Fri, 20 Mar 2020 02:00:27 +0000 (22:00 -0400)]
efi/gop: Allow specifying depth as well as resolution

Extend the video mode argument to handle an optional color depth
specification of the form
video=efifb:<xres>x<yres>[-(rgb|bgr|<bpp>)]

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-14-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Allow specifying mode by <xres>x<yres>
Arvind Sankar [Fri, 20 Mar 2020 02:00:26 +0000 (22:00 -0400)]
efi/gop: Allow specifying mode by <xres>x<yres>

Add the ability to choose a video mode using a command-line argument of
the form
video=efifb:<xres>x<yres>

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-13-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Allow specifying mode number on command line
Arvind Sankar [Fri, 20 Mar 2020 02:00:25 +0000 (22:00 -0400)]
efi/gop: Allow specifying mode number on command line

Add the ability to choose a video mode for the selected gop by using a
command-line argument of the form
video=efifb:mode=<n>

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-12-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Add prototypes for query_mode and set_mode
Arvind Sankar [Fri, 20 Mar 2020 02:00:24 +0000 (22:00 -0400)]
efi/gop: Add prototypes for query_mode and set_mode

Add prototypes and argmap for the Graphics Output Protocol's QueryMode
and SetMode functions.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-11-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Remove unreachable code from setup_pixel_info
Arvind Sankar [Fri, 20 Mar 2020 02:00:23 +0000 (22:00 -0400)]
efi/gop: Remove unreachable code from setup_pixel_info

pixel_format must be one of
PIXEL_RGB_RESERVED_8BIT_PER_COLOR
PIXEL_BGR_RESERVED_8BIT_PER_COLOR
PIXEL_BIT_MASK
since we skip PIXEL_BLT_ONLY when finding a gop.

Remove the redundant code and add another check in find_gop to skip any
pixel formats that we don't know about, in case a later version of the
UEFI spec adds one.

Reformat the code a little.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-10-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Use helper macros for find_bits
Arvind Sankar [Fri, 20 Mar 2020 02:00:22 +0000 (22:00 -0400)]
efi/gop: Use helper macros for find_bits

Use the __ffs/__fls macros to calculate the position and size of the
mask.

Correct type of mask to u32 instead of unsigned long.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-9-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Use helper macros for populating lfb_base
Arvind Sankar [Fri, 20 Mar 2020 02:00:21 +0000 (22:00 -0400)]
efi/gop: Use helper macros for populating lfb_base

Use the lower/upper_32_bits macros from kernel.h to initialize
si->lfb_base and si->ext_lfb_base.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-8-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Move variable declarations into loop block
Arvind Sankar [Fri, 20 Mar 2020 02:00:20 +0000 (22:00 -0400)]
efi/gop: Move variable declarations into loop block

Declare the variables inside the block where they're used.

Get rid of a couple of redundant initializers.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-7-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Slightly re-arrange logic of find_gop
Arvind Sankar [Fri, 20 Mar 2020 02:00:19 +0000 (22:00 -0400)]
efi/gop: Slightly re-arrange logic of find_gop

Small cleanup to get rid of conout_found.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-6-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Factor out locating the gop into a function
Arvind Sankar [Fri, 20 Mar 2020 02:00:18 +0000 (22:00 -0400)]
efi/gop: Factor out locating the gop into a function

Move the loop to find a gop into its own function.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-5-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Get mode information outside the loop
Arvind Sankar [Fri, 20 Mar 2020 02:00:17 +0000 (22:00 -0400)]
efi/gop: Get mode information outside the loop

Move extraction of the mode information parameters outside the loop to
find the gop, and eliminate some redundant variables.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-4-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Move check for framebuffer before con_out
Arvind Sankar [Fri, 20 Mar 2020 02:00:16 +0000 (22:00 -0400)]
efi/gop: Move check for framebuffer before con_out

If the gop doesn't have a framebuffer, there's no point in checking for
con_out support.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-3-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/gop: Remove redundant current_fb_base
Arvind Sankar [Fri, 20 Mar 2020 02:00:15 +0000 (22:00 -0400)]
efi/gop: Remove redundant current_fb_base

current_fb_base isn't used for anything except assigning to fb_base if
we locate a suitable gop.

Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
Link: https://lore.kernel.org/r/20200320020028.1936003-2-nivedita@alum.mit.edu
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub/arm: Make install_memreserve_table static
Zou Wei [Thu, 23 Apr 2020 12:08:33 +0000 (20:08 +0800)]
efi/libstub/arm: Make install_memreserve_table static

Fix the following sparse warning:

drivers/firmware/efi/libstub/arm-stub.c:68:6: warning:
symbol 'install_memreserve_table' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zou Wei <zou_wei@huawei.com>
Link: https://lore.kernel.org/r/1587643713-28169-1-git-send-email-zou_wei@huawei.com
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: unify EFI call wrappers for non-x86
Ard Biesheuvel [Thu, 23 Apr 2020 11:44:50 +0000 (13:44 +0200)]
efi/libstub: unify EFI call wrappers for non-x86

We have wrappers around EFI calls so that x86 can define special
versions for mixed mode, while all other architectures can use the
same simple definition that just issues the call directly.
In preparation for the arrival of yet another architecture that doesn't
need anything special here (RISC-V), let's move the default definition
into a shared header.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Make initrd file loader configurable
Ard Biesheuvel [Tue, 21 Apr 2020 07:17:35 +0000 (09:17 +0200)]
efi/libstub: Make initrd file loader configurable

Loading an initrd passed via the kernel command line is deprecated: it
is limited to files that reside in the same volume as the one the kernel
itself was loaded from, and we have more flexible ways to achieve the
same. So make it configurable so new architectures can decide not to
enable it.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoefi/libstub: Move arm-stub to a common file
Atish Patra [Wed, 15 Apr 2020 19:54:18 +0000 (12:54 -0700)]
efi/libstub: Move arm-stub to a common file

Most of the arm-stub code is written in an architecture independent manner.
As a result, RISC-V can reuse most of the arm-stub code.

Rename the arm-stub.c to efi-stub.c so that ARM, ARM64 and RISC-V can use it.
This patch doesn't introduce any functional changes.

Signed-off-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Palmer Dabbelt <palmerdabbelt@google.com>
Link: https://lore.kernel.org/r/20200415195422.19866-2-atish.patra@wdc.com
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
4 years agoLinux 5.7-rc2
Linus Torvalds [Sun, 19 Apr 2020 21:35:30 +0000 (14:35 -0700)]
Linux 5.7-rc2

4 years agomm: Fix MREMAP_DONTUNMAP accounting on VMA merge
Brian Geffon [Fri, 17 Apr 2020 17:25:56 +0000 (10:25 -0700)]
mm: Fix MREMAP_DONTUNMAP accounting on VMA merge

When remapping a mapping where a portion of a VMA is remapped
into another portion of the VMA it can cause the VMA to become
split. During the copy_vma operation the VMA can actually
be remerged if it's an anonymous VMA whose pages have not yet
been faulted. This isn't normally a problem because at the end
of the remap the original portion is unmapped causing it to
become split again.

However, MREMAP_DONTUNMAP leaves that original portion in place which
means that the VMA which was split and then remerged is not actually
split at the end of the mremap. This patch fixes a bug where
we don't detect that the VMAs got remerged and we end up
putting back VM_ACCOUNT on the next mapping which is completely
unreleated. When that next mapping is unmapped it results in
incorrectly unaccounting for the memory which was never accounted,
and eventually we will underflow on the memory comittment.

There is also another issue which is similar, we're currently
accouting for the number of pages in the new_vma but that's wrong.
We need to account for the length of the remap operation as that's
all that is being added. If there was a mapping already at that
location its comittment would have been adjusted as part of
the munmap at the start of the mremap.

A really simple repro can be seen in:
https://gist.github.com/bgaff/e101ce99da7d9a8c60acc641d07f312c

Fixes: e346b3813067 ("mm/mremap: add MREMAP_DONTUNMAP to mremap()")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Brian Geffon <bgeffon@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 20:59:06 +0000 (13:59 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "Two build fixes for a couple clk drivers and a fix for the Unisoc
  serial clk where we want to keep it on for earlycon"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: sprd: don't gate uart console clock
  clk: mmp2: fix link error without mmp2
  clk: asm9260: fix __clk_hw_register_fixed_rate_with_accuracy typo

4 years agoMerge tag 'x86-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 18:58:32 +0000 (11:58 -0700)]
Merge tag 'x86-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull x86 and objtool fixes from Thomas Gleixner:
 "A set of fixes for x86 and objtool:

  objtool:

   - Ignore the double UD2 which is emitted in BUG() when
     CONFIG_UBSAN_TRAP is enabled.

   - Support clang non-section symbols in objtool ORC dump

   - Fix switch table detection in .text.unlikely

   - Make the BP scratch register warning more robust.

  x86:

   - Increase microcode maximum patch size for AMD to cope with new CPUs
     which have a larger patch size.

   - Fix a crash in the resource control filesystem when the removal of
     the default resource group is attempted.

   - Preserve Code and Data Prioritization enabled state accross CPU
     hotplug.

   - Update split lock cpu matching to use the new X86_MATCH macros.

   - Change the split lock enumeration as Intel finaly decided that the
     IA32_CORE_CAPABILITIES bits are not architectural contrary to what
     the SDM claims. !@#%$^!

   - Add Tremont CPU models to the split lock detection cpu match.

   - Add a missing static attribute to make sparse happy"

* tag 'x86-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/split_lock: Add Tremont family CPU models
  x86/split_lock: Bits in IA32_CORE_CAPABILITIES are not architectural
  x86/resctrl: Preserve CDP enable over CPU hotplug
  x86/resctrl: Fix invalid attempt at removing the default resource group
  x86/split_lock: Update to use X86_MATCH_INTEL_FAM6_MODEL()
  x86/umip: Make umip_insns static
  x86/microcode/AMD: Increase microcode PATCH_MAX_SIZE
  objtool: Make BP scratch register warning more robust
  objtool: Fix switch table detection in .text.unlikely
  objtool: Support Clang non-section symbols in ORC generation
  objtool: Support Clang non-section symbols in ORC dump
  objtool: Fix CONFIG_UBSAN_TRAP unreachable warnings

4 years agoMerge tag 'timers-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Apr 2020 18:46:21 +0000 (11:46 -0700)]
Merge tag 'timers-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull time namespace fix from Thomas Gleixner:
 "An update for the proc interface of time namespaces: Use symbolic
  names instead of clockid numbers. The usability nuisance of numbers
  was noticed by Michael when polishing the man page"

* tag 'timers-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  proc, time/namespace: Show clock symbolic names in /proc/pid/timens_offsets

4 years agoMerge tag 'perf-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 18:28:01 +0000 (11:28 -0700)]
Merge tag 'perf-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull perf tooling fixes and updates from Thomas Gleixner:

 - Fix the header line of perf stat output for '--metric-only --per-socket'

 - Fix the python build with clang

 - The usual tools UAPI header synchronization

* tag 'perf-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tools headers: Synchronize linux/bits.h with the kernel sources
  tools headers: Adopt verbatim copy of compiletime_assert() from kernel sources
  tools headers: Update x86's syscall_64.tbl with the kernel sources
  tools headers UAPI: Sync drm/i915_drm.h with the kernel sources
  tools headers UAPI: Update tools's copy of drm.h headers
  tools headers kvm: Sync linux/kvm.h with the kernel sources
  tools headers UAPI: Sync linux/fscrypt.h with the kernel sources
  tools include UAPI: Sync linux/vhost.h with the kernel sources
  tools arch x86: Sync asm/cpufeatures.h with the kernel sources
  tools headers UAPI: Sync linux/mman.h with the kernel
  tools headers UAPI: Sync sched.h with the kernel
  tools headers: Update linux/vdso.h and grab a copy of vdso/const.h
  perf stat: Fix no metric header if --per-socket and --metric-only set
  perf python: Check if clang supports -fno-semantic-interposition
  tools arch x86: Sync the msr-index.h copy with the kernel sources

4 years agoMerge tag 'irq-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 18:23:33 +0000 (11:23 -0700)]
Merge tag 'irq-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "A set of fixes/updates for the interrupt subsystem:

   - Remove setup_irq() and remove_irq(). All users have been converted
     so remove them before new users surface.

   - A set of bugfixes for various interrupt chip drivers

   - Add a few missing static attributes to address sparse warnings"

* tag 'irq-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/irq-bcm7038-l1: Make bcm7038_l1_of_init() static
  irqchip/irq-mvebu-icu: Make legacy_bindings static
  irqchip/meson-gpio: Fix HARDIRQ-safe -> HARDIRQ-unsafe lock order
  irqchip/sifive-plic: Fix maximum priority threshold value
  irqchip/ti-sci-inta: Fix processing of masked irqs
  irqchip/mbigen: Free msi_desc on device teardown
  irqchip/gic-v4.1: Update effective affinity of virtual SGIs
  irqchip/gic-v4.1: Add support for VPENDBASER's Dirty+Valid signaling
  genirq: Remove setup_irq() and remove_irq()

4 years agoMerge tag 'sched-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Apr 2020 18:18:20 +0000 (11:18 -0700)]
Merge tag 'sched-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Thomas Gleixner:
 "Two fixes for the scheduler:

   - Work around an uninitialized variable warning where GCC can't
     figure it out.

   - Allow 'isolcpus=' to skip unknown subparameters so that older
     kernels work with the commandline of a newer kernel. Improve the
     error output while at it"

* tag 'sched-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/vtime: Work around an unitialized variable warning
  sched/isolation: Allow "isolcpus=" to skip unknown sub-parameters

4 years agoMerge tag 'core-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 18:16:00 +0000 (11:16 -0700)]
Merge tag 'core-urgent-2020-04-19' of git://git./linux/kernel/git/tip/tip

Pull RCU fix from Thomas Gleixner:
 "A single bugfix for RCU to prevent taking a lock in NMI context"

* tag 'core-urgent-2020-04-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rcu: Don't acquire lock in NMI handler in rcu_nmi_enter_common()

4 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Apr 2020 18:05:15 +0000 (11:05 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Miscellaneous bug fixes and cleanups for ext4, including a fix for
  generic/388 in data=journal mode, removing some BUG_ON's, and cleaning
  up some compiler warnings"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: convert BUG_ON's to WARN_ON's in mballoc.c
  ext4: increase wait time needed before reuse of deleted inode numbers
  ext4: remove set but not used variable 'es' in ext4_jbd2.c
  ext4: remove set but not used variable 'es'
  ext4: do not zeroout extents beyond i_disksize
  ext4: fix return-value types in several function comments
  ext4: use non-movable memory for superblock readahead
  ext4: use matching invalidatepage in ext4_writepage

4 years agoMerge tag '5.7-rc-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 19 Apr 2020 18:00:27 +0000 (11:00 -0700)]
Merge tag '5.7-rc-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Three small smb3 fixes: two debug related (helping network tracing for
  SMB2 mounts, and the other removing an unintended debug line on
  signing failures), and one fixing a performance problem with 64K
  pages"

* tag '5.7-rc-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: remove overly noisy debug line in signing errors
  cifs: improve read performance for page size 64KB & cache=strict & vers=2.1+
  cifs: dump the session id and keys also for SMB2 sessions

4 years agoMerge tag 'flexible-array-member-5.7-rc2' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Sun, 19 Apr 2020 17:34:30 +0000 (10:34 -0700)]
Merge tag 'flexible-array-member-5.7-rc2' of git://git./linux/kernel/git/gustavoars/linux

Pull flexible-array member conversion from Gustavo Silva:
 "The current codebase makes use of the zero-length array language
  extension to the C90 standard, but the preferred mechanism to declare
  variable-length types such as these ones is a flexible array
  member[1][2], introduced in C99:

    struct foo {
        int stuff;
        struct boo array[];
    };

  By making use of the mechanism above, we will get a compiler warning
  in case the flexible array does not occur last in the structure, which
  will help us prevent some kind of undefined behavior bugs from being
  inadvertently introduced[3] to the codebase from now on.

  Also, notice that, dynamic memory allocations won't be affected by
  this change:

   "Flexible array members have incomplete type, and so the sizeof
    operator may not be applied. As a quirk of the original
    implementation of zero-length arrays, sizeof evaluates to zero."[1]

  sizeof(flexible-array-member) triggers a warning because flexible
  array members have incomplete type[1]. There are some instances of
  code in which the sizeof operator is being incorrectly/erroneously
  applied to zero-length arrays and the result is zero. Such instances
  may be hiding some bugs. So, this work (flexible-array member
  convertions) will also help to get completely rid of those sorts of
  issues.

  Notice that all of these patches have been baking in linux-next for
  quite a while now and, 238 more of these patches have already been
  merged into 5.7-rc1.

  There are a couple hundred more of these issues waiting to be
  addressed in the whole codebase"

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

* tag 'flexible-array-member-5.7-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux: (28 commits)
  xattr.h: Replace zero-length array with flexible-array member
  uapi: linux: fiemap.h: Replace zero-length array with flexible-array member
  uapi: linux: dlm_device.h: Replace zero-length array with flexible-array member
  tpm_eventlog.h: Replace zero-length array with flexible-array member
  ti_wilink_st.h: Replace zero-length array with flexible-array member
  swap.h: Replace zero-length array with flexible-array member
  skbuff.h: Replace zero-length array with flexible-array member
  sched: topology.h: Replace zero-length array with flexible-array member
  rslib.h: Replace zero-length array with flexible-array member
  rio.h: Replace zero-length array with flexible-array member
  posix_acl.h: Replace zero-length array with flexible-array member
  platform_data: wilco-ec.h: Replace zero-length array with flexible-array member
  memcontrol.h: Replace zero-length array with flexible-array member
  list_lru.h: Replace zero-length array with flexible-array member
  lib: cpu_rmap: Replace zero-length array with flexible-array member
  irq.h: Replace zero-length array with flexible-array member
  ihex.h: Replace zero-length array with flexible-array member
  igmp.h: Replace zero-length array with flexible-array member
  genalloc.h: Replace zero-length array with flexible-array member
  ethtool.h: Replace zero-length array with flexible-array member
  ...

4 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 18 Apr 2020 21:03:12 +0000 (14:03 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Seven fixes: three in target, one on a sg error leg, two in qla2xxx
  fixing warnings introduced in the last merge window and updating
  MAINTAINERS and one in hisi_sas fixing a problem introduced by libata"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: sg: add sg_remove_request in sg_common_write
  scsi: target: tcmu: reset_ring should reset TCMU_DEV_BIT_BROKEN
  scsi: target: fix PR IN / READ FULL STATUS for FC
  scsi: target: Write NULL to *port_nexus_ptr if no ISID
  scsi: MAINTAINERS: Update qla2xxx FC-SCSI driver maintainer
  scsi: qla2xxx: Fix regression warnings
  scsi: hisi_sas: Fix build error without SATA_HOST

4 years agoxattr.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:41:14 +0000 (19:41 -0500)]
xattr.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agouapi: linux: fiemap.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 02:30:22 +0000 (21:30 -0500)]
uapi: linux: fiemap.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agouapi: linux: dlm_device.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:51:46 +0000 (19:51 -0500)]
uapi: linux: dlm_device.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agotpm_eventlog.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:38:18 +0000 (19:38 -0500)]
tpm_eventlog.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agoti_wilink_st.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:25:06 +0000 (19:25 -0500)]
ti_wilink_st.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agoswap.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:23:10 +0000 (19:23 -0500)]
swap.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agoskbuff.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:22:24 +0000 (19:22 -0500)]
skbuff.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agosched: topology.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:14:37 +0000 (19:14 -0500)]
sched: topology.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agorslib.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:13:20 +0000 (19:13 -0500)]
rslib.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
4 years agorio.h: Replace zero-length array with flexible-array member
Gustavo A. R. Silva [Tue, 24 Mar 2020 00:12:17 +0000 (19:12 -0500)]
rio.h: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>